A recurrent neural network based method for predicting the state of aircraft air conditioning system

The reliability and safety of aircraft has always been the focus of research attention. As an important component of the aircraft, the air conditioning system has a direct impact on the safety of the flight process. In order to ensure the safety of the flight process, this paper proposes a recurrent neural network (RNN) based method for predicting the working state of the aircraft air conditioning system. Using the measured data collected from the Boeing 737NG aircraft, we train a RNN and experimental results on short-term prediction show that our proposed method can obtain a good prediction accuracy. In addition, we modify the network to make longer prediction using a bidirectional architecture. The experimental results on long-term prediction show that this network can solve the problem that the prediction results at first several seconds are much larger than the actual measured value and can learn a good representation for the time series.